[
https://issues.apache.org/jira/browse/HADOOP-10942?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14093535#comment-14093535
]
Colin Patrick McCabe commented on HADOOP-10942:
-----------------------------------------------
It seems like there's a bunch of things going on here:
* The globber will sometimes erroneously return a permission denied exception
when there is a non-terminal wildcard. For example, when listing {{/a*/b}}, if
there is a file named /alpha, the glob will fail. This bug does *not* occur
for superusers, which is why the existing tests and casual testing didn't catch
it. You mention that this was a "fix" which was regressed between 0.23 and
branch-2... is there a JIRA number for this already?
* Optimizations: you mention "doing a simple immediate file status if the path
contains no globs, etc". The existing code already does this. It was added in
HADOOP-9877. Are we missing a case? I didn't understand the comment about
"Additional optimizations to reduce rpcs actually increases them if directories
contain 1 item." Which specific optimization(s) are increasing RPCs for you
and how can we avoid this?
* You added a comment that "FileContext returns a path to the home dir of the
user that started the jvm instead of the ugi user so we'll just workaround it."
I wasn't aware of this issue. Is there a JIRA number? This seems like an
inconsistency that we should note in the test, along with a link to the JIRA
that should fix it.
* There's a bunch of reorganization here, perhaps almost a rewrite of the main
part of the globber.
Let's split these into separate JIRAs so that it's easier to review.
> Globbing optimizations and regression fix
> -----------------------------------------
>
> Key: HADOOP-10942
> URL: https://issues.apache.org/jira/browse/HADOOP-10942
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs
> Affects Versions: 3.0.0, 2.1.0-beta
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
> Priority: Critical
> Attachments: HADOOP-10942.patch
>
>
> When globbing was commonized to support both filesystem and filecontext, it
> regressed a fix that prevents an intermediate glob that matches a file from
> throwing a confusing permissions exception. The hdfs traverse check requires
> the exec bit which a file does not have.
> Additional optimizations to reduce rpcs actually increases them if
> directories contain 1 item.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
